Advice needed: Is it acceptable for time-frequency plot to show significant activity but not for GFP?

Below are two example pairs of time-frequency and the corresponding GFP plots for the orange v. blue conditions. Data were calculated over the mean activity over 12 sensors.

I calculated the mean GFP by first obtaining the mean FT/cm within the dark greyed window time frame for each of the 12 sensors, then calculated the average (of the 12 sensors) of the averaged time windows. When running a t-test, no significant difference was found between the GFP activity for orange v. blue. I understand that the way the data were structured and the analysis pipeline for time-frequency analysis and GFP are different, so it might not be surprising that significance in one analysis might not translate to the other. However, in MEG papers that I have seen, it is common for GFP plot to show a much more prominent peak at the time window of interests that often corresponds with the time-frequency result.

My three questions here are,
1) Do GFP activity sometimes not capture significant activity seen in time-frequency analysis?
2) Would it be acceptable in peer review standards to have GFP activity not be consistent with the time-frequency output?
3) In my cases here, does it mean that there are low firing, but consistent sustained activity over time going in the alpha-beta range as opposed to the more typical event related single evoked activity spike?

Appreciate any advice here!

To me your time traces look like they came from the result of a Hilbert transformation, is that correct? Either way, if you take your time data, band-pass filter, hilbert transform and take the magnitude (i.e., envelope=True in our code), and then low-pass this result, what you’ll get will be more consistent. When you look at the oscillatory data, the peaks and valleys of the blue and yellow will make the raw GFP itself not show the same information as the TFR or Hilbert-based analysis I’m describing.

1 Like